1 # Copyright (C) 2015, all contributors <meta@public-inbox.org>
2 # License: AGPLv3 or later (https://www.gnu.org/licenses/agpl-3.0.txt)
3 # based on notmuch, but with no concept of folders, files or flags
4 package PublicInbox::Search;
7 use PublicInbox::SearchMsg;
8 use Search::Xapian qw/:standard/;
9 require PublicInbox::View;
11 use PublicInbox::MID qw/mid_clean mid_compressed/;
15 # SCHEMA_VERSION history
17 # 1 - subject_path is lower-cased
18 # 2 - subject_path is mid_compressed in the index, only
19 # 3 - message-ID is compressed if it includes '%' (hack!)
22 QP_FLAGS => FLAG_PHRASE|FLAG_BOOLEAN|FLAG_LOVEHATE|FLAG_WILDCARD,
26 my %bool_pfx_internal = (
27 type => 'T', # "mail" or "ghost"
28 mid => 'Q', # uniQue id (Message-ID or mid_compressed)
31 my %bool_pfx_external = (
33 thread => 'G', # newsGroup (or similar entity - e.g. a web forum name)
34 references => 'XREFS',
42 my %all_pfx = (%bool_pfx_internal, %bool_pfx_external, %prob_prefix);
44 sub xpfx { $all_pfx{$_[0]} }
47 while (my ($k, $v) = each %all_pfx) {
48 next if $prob_prefix{$k};
49 $PFX2TERM_RMAP{$v} = $k;
52 my $mail_query = Search::Xapian::Query->new(xpfx('type') . 'mail');
55 my ($class, $git_dir, $writable) = @_;
56 # allow concurrent versions for easier rollback:
57 my $dir = "$git_dir/public-inbox/xapian" . SCHEMA_VERSION;
60 if ($writable) { # not used by the WWW interface
61 require Search::Xapian::WritableDatabase;
63 File::Path::mkpath($dir);
64 $db = Search::Xapian::WritableDatabase->new($dir,
65 Search::Xapian::DB_CREATE_OR_OPEN);
67 $db = Search::Xapian::Database->new($dir);
69 bless { xdb => $db, git_dir => $git_dir }, $class;
72 sub reopen { $_[0]->{xdb}->reopen }
75 my ($self, $mime) = @_; # mime = Email::MIME object
76 my $db = $self->{xdb};
79 my $mid = mid_clean($mime->header_obj->header_raw('Message-ID'));
80 $mid = mid_compressed($mid);
82 my $ct_msg = $mime->header('Content-Type') || 'text/plain';
83 my $enc_msg = PublicInbox::View::enc_for($ct_msg);
85 $db->begin_transaction;
87 my $smsg = $self->lookup_message($mid);
91 $smsg->ensure_metadata;
92 # convert a ghost to a regular message
93 # it will also clobber any existing regular message
97 my $type = xpfx('type');
99 $doc->remove_term($type . 'ghost');
103 # probably does not exist:
104 eval { $doc->remove_term($type . 'mail') };
105 $doc->add_term($type . 'mail');
107 $smsg = PublicInbox::SearchMsg->new($mime);
109 $doc->add_term(xpfx('mid') . $mid);
112 my $subj = $smsg->subject;
115 $doc->add_term(xpfx('subject') . $subj);
117 my $path = subject_path($subj);
118 $doc->add_term(xpfx('path') . mid_compressed($path));
121 my $from = $smsg->from_name;
122 my $date = $smsg->date;
123 my $ts = Search::Xapian::sortable_serialise($smsg->ts);
124 $doc->add_value(PublicInbox::Search::TS, $ts);
126 $doc->set_data($smsg->to_doc_data);
128 my $tg = $self->term_generator;
130 $tg->set_document($doc);
131 $tg->index_text($subj, 1, 'S') if $subj;
132 $tg->increase_termpos;
133 $tg->index_text($subj) if $subj;
134 $tg->increase_termpos;
136 $tg->index_text($smsg->from->format);
137 $tg->increase_termpos;
139 $mime->walk_parts(sub {
141 return if $part->subparts; # walk_parts already recurses
142 my $ct = $part->content_type || $ct_msg;
144 # account for filter bugs...
145 $ct =~ m!\btext/plain\b!i or return;
147 my $enc = PublicInbox::View::enc_for($ct, $enc_msg);
149 foreach my $l (split(/\n/, $enc->decode($part->body))) {
157 $tg->index_text(join("\n", @quot), 0);
158 $tg->increase_termpos;
161 $tg->index_text(join("\n", @orig));
162 $tg->increase_termpos;
167 $doc_id = $smsg->doc_id;
168 $self->link_message($smsg, 0);
169 $db->replace_document($doc_id, $doc);
171 $self->link_message($smsg, 0);
172 $doc_id = $db->add_document($doc);
177 warn "failed to index message <$mid>: $@\n";
178 $db->cancel_transaction;
180 $db->commit_transaction;
185 # returns deleted doc_id on success, undef on missing
187 my ($self, $mid) = @_;
188 my $db = $self->{xdb};
190 $mid = mid_clean($mid);
191 $mid = mid_compressed($mid);
193 $db->begin_transaction;
195 $doc_id = $self->find_unique_doc_id('mid', $mid);
196 $db->delete_document($doc_id) if defined $doc_id;
200 warn "failed to remove message <$mid>: $@\n";
201 $db->cancel_transaction;
203 $db->commit_transaction;
210 my ($self, $query_string, $opts) = @_;
211 my $query = $self->qp->parse_query($query_string, QP_FLAGS);
213 $query = Search::Xapian::Query->new(OP_AND, $mail_query, $query);
214 $self->do_enquire($query, $opts);
217 sub get_subject_path {
218 my ($self, $path, $opts) = @_;
219 my $query = $self->qp->parse_query("path:".mid_compressed($path), 0);
220 $self->do_enquire($query);
223 # given a message ID, get followups to a message
225 my ($self, $mid, $opts) = @_;
226 $mid = mid_clean($mid);
227 $mid = mid_compressed($mid);
229 my $irt = $qp->parse_query("inreplyto:$mid", 0);
230 my $ref = $qp->parse_query("references:$mid", 0);
231 my $query = Search::Xapian::Query->new(OP_OR, $irt, $ref);
233 $self->do_enquire($query);
237 my ($self, $mid, $opts) = @_;
238 my $smsg = eval { $self->lookup_message($mid) };
240 return { count => 0, msgs => [] } unless $smsg;
242 my $qtid = $qp->parse_query('thread:'.$smsg->thread_id);
243 my $qsub = $qp->parse_query('path:'.mid_compressed($smsg->path));
244 my $query = Search::Xapian::Query->new(OP_OR, $qtid, $qsub);
245 $self->do_enquire($query);
251 my ($self, $query, $opts) = @_;
252 my $enquire = $self->enquire;
254 $enquire->set_query($query);
255 $enquire->set_sort_by_relevance_then_value(TS, 0);
257 my $offset = $opts->{offset} || 0;
258 my $limit = $opts->{limit} || 50;
259 my $mset = $enquire->get_mset($offset, $limit);
261 PublicInbox::SearchMsg->load_doc($_->get_document);
264 { count => $mset->get_matches_estimated, msgs => \@msgs }
268 sub stemmer { Search::Xapian::Stem->new(LANG) }
274 my $qp = $self->{query_parser};
278 $qp = Search::Xapian::QueryParser->new;
279 $qp->set_default_op(OP_AND);
280 $qp->set_database($self->{xdb});
281 $qp->set_stemmer($self->stemmer);
282 $qp->set_stemming_strategy(STEM_SOME);
283 $qp->add_valuerangeprocessor($self->ts_range_processor);
284 $qp->add_valuerangeprocessor($self->date_range_processor);
286 while (my ($name, $prefix) = each %bool_pfx_external) {
287 $qp->add_boolean_prefix($name, $prefix);
290 while (my ($name, $prefix) = each %prob_prefix) {
291 $qp->add_prefix($name, $prefix);
294 $self->{query_parser} = $qp;
297 sub term_generator { # write-only
300 my $tg = $self->{term_generator};
303 $tg = Search::Xapian::TermGenerator->new;
304 $tg->set_stemmer($self->stemmer);
306 $self->{term_generator} = $tg;
309 sub next_doc_id { $_[0]->{xdb}->get_lastdocid + 1 }
311 # increments last_thread_id counter
312 # returns a 64-bit integer represented as a hex string
315 my $db = $self->{xdb};
316 my $last_thread_id = int($db->get_metadata('last_thread_id') || 0);
318 $db->set_metadata('last_thread_id', ++$last_thread_id);
323 sub ts_range_processor {
324 $_[0]->{tsrp} ||= Search::Xapian::NumberValueRangeProcessor->new(TS);
327 sub date_range_processor {
328 $_[0]->{drp} ||= Search::Xapian::DateValueRangeProcessor->new(TS);
332 my ($self, $smsg, $is_ghost) = @_;
335 $smsg->ensure_metadata;
337 $self->link_message_to_parents($smsg);
341 sub link_message_to_parents {
342 my ($self, $smsg) = @_;
343 my $doc = $smsg->{doc};
344 my $mid = mid_compressed($smsg->mid);
345 my $mime = $smsg->mime;
346 my $refs = $mime->header_obj->header_raw('References');
347 my @refs = $refs ? ($refs =~ /<([^>]+)>/g) : ();
348 my $irt = $mime->header_obj->header_raw('In-Reply-To');
350 if ($irt =~ /<([^>]+)>/) {
358 @refs = map { mid_compressed($_) } @refs;
360 @refs = grep { !$uniq{$_}++ } @refs; # uniq
362 $doc->add_term(xpfx('inreplyto') . $refs[-1]);
364 my $ref_pfx = xpfx('references');
366 # first ref *should* be the thread root,
367 # but we can never trust clients to do the right thing
368 my $ref = shift @refs;
369 $doc->add_term($ref_pfx . $ref);
370 $tid = $self->_resolve_mid_to_tid($ref);
372 # the rest of the refs should point to this tid:
373 foreach $ref (@refs) {
374 $doc->add_term($ref_pfx . $ref);
375 my $ptid = $self->_resolve_mid_to_tid($ref);
377 $self->merge_threads($tid, $ptid);
381 $tid = $self->next_thread_id;
383 $doc->add_term(xpfx('thread') . $tid);
387 my ($self, $mid) = @_;
388 $mid = mid_clean($mid);
389 $mid = mid_compressed($mid);
391 my $doc_id = $self->find_unique_doc_id('mid', $mid);
393 if (defined $doc_id) {
395 my $doc = $self->{xdb}->get_document($doc_id);
396 $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
397 $smsg->doc_id($doc_id);
402 sub find_unique_doc_id {
403 my ($self, $term, $value) = @_;
405 my ($begin, $end) = $self->find_doc_ids($term, $value);
407 return undef if $begin->equal($end); # not found
409 my $rv = $begin->get_docid;
413 $begin->equal($end) or die "Term '$term:$value' is not unique\n";
417 # returns begin and end PostingIterator
419 my ($self, $term, $value) = @_;
421 $self->find_doc_ids_for_term(xpfx($term) . $value);
424 # returns begin and end PostingIterator
425 sub find_doc_ids_for_term {
426 my ($self, $term) = @_;
427 my $db = $self->{xdb};
429 ($db->postlist_begin($term), $db->postlist_end($term));
432 # this will create a ghost as necessary
433 sub _resolve_mid_to_tid {
434 my ($self, $mid) = @_;
436 my $smsg = $self->lookup_message($mid) || $self->create_ghost($mid);
441 my ($self, $mid, $tid) = @_;
443 $mid = mid_compressed($mid);
444 $tid = $self->next_thread_id unless defined $tid;
446 my $doc = Search::Xapian::Document->new;
447 $doc->add_term(xpfx('mid') . $mid);
448 $doc->add_term(xpfx('thread') . $tid);
449 $doc->add_term(xpfx('type') . 'ghost');
451 my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
452 $self->link_message($smsg, 1);
453 $self->{xdb}->add_document($doc);
459 my ($self, $winner_tid, $loser_tid) = @_;
460 my ($head, $tail) = $self->find_doc_ids('thread', $loser_tid);
461 my $thread_pfx = xpfx('thread');
462 my $db = $self->{xdb};
464 for (; $head != $tail; $head->inc) {
465 my $docid = $head->get_docid;
466 my $doc = $db->get_document($docid);
467 $doc->remove_term($thread_pfx . $loser_tid);
468 $doc->add_term($thread_pfx . $winner_tid);
469 $db->replace_document($docid, $doc);
473 # normalize subjects so they are suitable as pathnames for URLs
479 $subj =~ s/^(?:re|aw):\s*//i; # remove reply prefix (aw: German)
480 $subj =~ s![^a-zA-Z0-9_\.~/\-]+!_!g;
485 my ($git, $blob) = @_;
487 my $str = $git->cat_file($blob);
488 Email::MIME->new($str);
494 my ($self, $git, $blob) = @_;
495 my $mime = do_cat_mail($git, $blob) or return;
496 eval { $self->add_message($mime) };
497 warn "W: index_blob $blob: $@\n" if $@;
501 my ($self, $git, $blob) = @_;
502 my $mime = do_cat_mail($git, $blob) or return;
503 my $mid = $mime->header_obj->header_raw('Message-ID');
504 eval { $self->remove_message($mid) } if defined $mid;
505 warn "W: unindex_blob $blob: $@\n" if $@;
510 $self->{enquire} ||= Search::Xapian::Enquire->new($self->{xdb});
513 # indexes all unindexed messages
515 my ($self, $git) = @_;
516 my $db = $self->{xdb};
517 my $latest = $db->get_metadata('last_commit');
518 my $range = length $latest ? "$latest..HEAD" : 'HEAD';
521 my $hex = '[a-f0-9]';
522 my $h40 = $hex .'{40}';
523 my $addmsg = qr!^:000000 100644 \S+ ($h40) A\t${hex}{2}/${hex}{38}$!;
524 my $delmsg = qr!^:100644 000000 ($h40) \S+ D\t${hex}{2}/${hex}{38}$!;
526 # get indexed messages
527 my @cmd = ('git', "--git-dir=$git->{git_dir}", "log",
528 qw/--reverse --no-notes --no-color --raw -r --no-abbrev/,
531 my $pid = open(my $log, '-|', @cmd) or
532 die('open` '.join(' ', @cmd) . " pipe failed: $!\n");
534 while (my $line = <$log>) {
535 if ($line =~ /$addmsg/o) {
536 $self->index_blob($git, $1);
537 } elsif ($line =~ /$delmsg/o) {
538 $self->unindex_blob($git, $1);
539 } elsif ($line =~ /^commit ($h40)/o) {
541 if (defined $latest) {
542 $db->set_metadata('last_commit', $latest)
548 $db->set_metadata('last_commit', $latest) if defined $latest;